The rapid digitization and decentralization of modern energy systems have created vast amounts of heterogeneous data modeled with diverse ontologies. Ensuring semantic interoperability across these sources is essential but remains challenging due to domain-specific terminology and complex structural correspondences found in different ontologies. Traditional ontology matchers, which rely mainly on lexical or structural information, often struggle to capture the complex semantic. This paper presents a hybrid ontology matching framework designed for the energy domain. The approach combines a domain-pretrained transformer model with a large language model. Experiments on real-world energy ontologies show that the proposed method outperforms classical matchers, transformer-only approaches, and recent LLM-based systems. These results indicate that combining domain-specific encoders with prompt-guided LLM refinement offers an effective and scalable solution for ontology matching in a complex energy context.